Model-Based Decision Support for Industry-Environment Interactions

Working Paper

Model-based Decision Support for Industry-Environment Interactions A Pesticide Industry Example

K . Fcdra M . Karhu T . R y s M . SCocz M . Zebrowski

W . Zicmbla

November 1987 W P-87-97

International Institute for Applied Systems Analysis

A-2361 Laxenburg, Austria

Model-based Decision Support for Industry-Environment Interactions A Pesticide Industry Example

K . Fedra hi. Karhu T . Rys hi. Skocz hi. Zebrowski

W . Ziembla

November 1987 W P-87-97

Working Papers are interim reports on work of the International Institute for Applied Systems Analysis and have received only limited review. Views or opinions expressed herein do not necessarily represent those of the Institute or of its National Member Organizations.

INTERNATIONAL INSTITUTE FOR APPLIED SYSTEMS ANALYSIS A-2361 Laxenburg, Austria

PREFACE

Applied systems analysis is - or should be - a tool in the hands of planners and de- cision makers who have to deal with the complex and growing problems of modern so- ciety. There is, however, an obvious gap between the ever-increasing complexity and volume of scientific and technological information and tools of analysis relevant t o large socio-technical and environmental systems, and the information requirements a t a stra- tegic planning and policy level.

The Advanced Computer Applications (ACA) project builds on IIASA's traditional strength in the methodological foundations of operations research and applied systems analysis, and its rich experience in numerous application areas including the environ- ment, technology, and risk. The ACA group draws on this infrastructure and combines it with elements of A1 and advanced information and computer technology. Several completely externally-funded research and development projects in the field of model- based decision support and applied Artificial Intelligence (AT) are currently under way.

As an example of this approach to information and decision support systems, one of the components of an R & D project sponsored by the CEC's EURATOM Joint Research Centre (JRC) a t Ispra, Italy, in the area of hazardous substances and indus- trial risk management, is described in this paper. The PDA (Production Distribution Area) is an interactive optimization code (based on DIDASS, one of a family of multi- criteria decision support tools developed at IIASA) and a linear problem solver, for chemical industry structures, configured for the pesticide industry of a hypothetical region.

The user can select optimization criteria, define allowable ranges or constraints on these criteria, define reference points for the multi-criteria trade-off, and display various levels of model output, including the waste streams generated by the different industrial structure alternatives. These waste streams can then be used to provide input conditions for the environmental impact models.

With the emphasis on a directly understandable problem representation and dynamic color graphics, and the user interface as a k e y element of interactive decision support systems, this is a step toward increased direct practical usability of IIASA's research results.

Robert H. Pry Director

ACKNOWLEDGEMENTS

The research described in. this report is being carried out by IIASA's Advanced Computer Applications (ACA) project in cooperation with the Institute for Control and Systems Engineering, Academy of Mining and Metallurgy (ICSEAMM), Cracow, P e land. This cooperation is based on an agreement between IIASA and the Committee for Coordination of Cooperation with IlASA of the Polish Academy of Sciences (National Member Organization).

First components of the system described have been integrated in software developed under contract to the Commission of the European Communities' (CEC) Joint Research Centre (JRC), Ispra Establishment, under Study Contracts No.2524-84-

11 ED ISP A and 2748-8507 ED ISP A.

The opinions expressed in the report are those of the authors and do not necessarily reflect those of IIASA or of IIASA's National Member Organizations. Neither the colla- borating institutes, the Commission of the European Communities, the Joint Research Centre, Ispra Establishment, nor any person acting on behalf of the above is responsible for the use which might be made of the information in this report.

T m AUTHORS

Kurt Fedra, Austria, is the leader of IIASA's Advanced Computer Applications (ACA) project.

Marrku Karhu, Finland, from the Technical Research Center of Finland (VTT), was a participant in the 1986 Young Summer Scientists Program (YSSP) at IIASA.

Maciej Zebrowski, Poland, is Deputy Director of the Institute for Control and Sys- - terns Engineering, Academy of Mining and Metallurgy (ICSEAMM), Cracow.

T.Rys, M-Skocz, and W.Ziembla are staff scientists a t ICSEAMM, Cracow.

-

-

CONTENTS

1. The Problem Context 1

1.1 The Project Background 2

1.2 A Summary Description of the Optimization Model 3

1.3 The Pesticide Industry Case Study 5

2. A Guided Tour through the Model System 8

2.1 Getting Started 8

2.2 Databases and Background Information 9

2.2.1 Production Technologies 9

2.2.2 Process Waste Streams 10

2.2.3 Industrial Establishments Database 11

2.2.4 Hazardous Substances Database 11

2.3 Numerical Experiments: Interactive Optimization 13 2.3.1 Defining Scenarios for Optimization 15

2.3.2 Running the Problem Solver 18

2.4 Model Output and Evaluation 18

2.4.1 Results by Technology 19

2.4.2 Listing of Waste Stream Constituents 2 1

2.5 Environmental Impact Simulation 22

2.6 Scenario Evaluation: a Discrete Post-processor 23 3. Basic Concepts underlying the DSS Development 26 3.1 PDA Model Formulation

3.2 The Linear Programming Probleni 3.3 The Iliscrete Post-processor 3.4 Implementation

4. References 30

Model-based Decision Support for Industry-Environment Interactions

A Pesticide Industry Example

K . Fedra, M . Karhu, T . Rys, M . Skocz, M . Zebrowski, W . Ziembla

1. The Problem Context

Whether they appear as raw materials, as finished products, as by-products, or as wastes, hazardous substances pose risks t o man and the environment which must be responsibly managed.

The annual waste generated in the countries of the EC amounts t o about 2 giga- tons. Somewhat less than 10% of this colossal amount is of industrial origin. Roughly 10% of these industrial wastes have t o be classified as hazardous (J. Schneider, JRC/Ispra, 1984, personal communication). More graphically, this hazardous waste production amounts t o approximately 20 million metric tons, that could fill a train of roughly 10,000 km in length.

The effective management of these wastes requires:

a minimization of waste production by process modification and recycling;

the conversion t o non-hazardous forms, i-e., treatment;

finally, a safe disposal of whatever is left.

In addition t o hazardous wastes, there is a large number of commercial products t h a t must be considered hazardous. This is particularly true for the case of the pesticide industry discussed here.

The regulatory framework for hazardous substances within the European Commun- ity is largely defined by a number of Directives of the Council of the European Commun- ities and the corresponding national legislation which these Directives require (see, e.g., Haigh, 1984; Majone, 1985; Baram, 1985). For example, the so-called Seveso Directive (Council Directive on the major accident hazards of certain industrial activities, 82/501/EEC) specifies that manufacturers must provide the competent authorities with information on the details of substances and processes involved in high-risk facilities.

Further, people outside the establishment who might be affected by a major accident must be informed of the safety measures t o be taken in the event of an emergency.

The Council Directive on toxic and dangerous wastes (78/319/EEC) calls for a comprehensive system of monitoring and supervision of facilities and operations involv- ing hazardous wastes, specifically mentioning risks t o water, air, soil, plants and animals, while also including nuisance due t o noise and odors and possible degradation of the countryside and places of special interest. More recently, the Directive on the assessment of the effects of certain public and private projects on the environment (85/337/EEC, June 1985) requires comprehensive environmental assessments of projects

a and installations involving hazardous materials. These assessments are t o include con- sideration of the production and storage of materials such as pesticides, pharmaceuticals, paints, etc. A broad analysis of the direct and indirect effects on people, environment,

property and cultural heritage is also foreseen and the evaluation of alternatives is re- quired. A more detailed discussion of the regulatory and institutional framework and the role computers might play in it as part of the decision-making process, is found in Fedra and Otway (1986) and Otway and Peltu (1985).

Obviously, regulation of hazardous substances and industrial risks is most effective a t the source, i.e., a t the industrial production process level. Options for the regulator range from the outright ban of specific products to imposing constraints (direct regulate ry or possibly indirect monetary through taxes and fees) on production facilities and waste treatment and disposal options.

The model-based decision support system described here is designed to address problems that arise in the course of, for example, the analysis of regulatory options.

The central component is an optimization model that describes the behavior of a chemi- cal industry, given certain assumptions about prices for products raw materials, and la- bor, upper and lower limits for certain production lines or waste products, under the basic assumption that the industry will operate t o maximize its net economic results while meeting the external constraints. The results of changes in these external condi- tions (reflecting the market as well as a set of regulatory options) will be a redistribution of production capacities, resulting in a different product mix with different effects on the environment.

In other words, the model will show what a rational industry might do given a cer- tain set of regulations under specific market conditions. It may be worthwhile noting that the market itself is not included in the model; prices are fixed and set externally, i.e., by the user, and an adjustment of production volumes does not (within the model) afiect prices.

The representation of economics in the model is certainly very simplistic, in part constrained by the linear model used. The major advantage of the model, however, is its fast and reliable bookkeeping of albeit simplified material flows and basic cost com- ponents, that allow a fast and interactive screening of regulatory options. Auxiliary da- tabases, a conversational control over display options, coupled environmental impact analysis that translate waste streams generated directly into environmental quality indi- cators such as water quality, and finally a post-processor for the comparative evaluation of several optimization experiments integrate into a very powerful, but easy-teuse software tool.

1.1 The Project Background

An early prototype of the model systern described here was integrated as part of an integrated software system for the management of hazardous substances (Fedra, 1985;

1986; Fedra and Otway 1986) * ) for the analysis of the chemical industry, the simulation of its behavior and optimization of its structure.

The central optimization model is implemented on the basis of the PDA model (Dobrowolski et al., 1982, 1984; Zebrowski et al., 1985) and the relevant MIDA m e t h e dology (Dobrowolski et al., 1985). For the pesticide industry example, the modified ver- sion of the model and corresponding software implementation was developed.**)

The aim of the overall project is to provide software tools which can be used by those engaged in the management of the environment, industrial production, products, and waste streams, and hazardous substances and wastes in particular. This set of tools his software system was developed by IIASAYs Advanced Computer Applications (ACA) Project,

~ ~ d e r contract t o t h e CEC's Joint Research Centre (JIXC). I ~ p r a , Italy

his model and eoftware was developed by the Joint Systems Research Department (JSRD) of t h e

m Academy of Mining and Metallurgy, C r n c o ~ . T')land, under contract to IIASA and in collaboration with t h e Advanced Computer Applicatione ( A C A ) Project.

is designed for a broad group of users, including non-technical users. Its primary pur- pose is to improve the factual basis for deciaion making, and to structure the deciaion- making process in order to make it more consistent, by providing easy access and allow- ing efficient use of methods of analysis and information management which are normally restricted to a small group of technical experts.

In order t o design and develop an integrated eet of eojtware tools, we build on exist- ing models and computer-assisted procedures. For the casual user, and for more experi- mental and explorative use, it also appears necessary to build much of the accumulated knowledge of the subject areas into the user interface for the models. Thus, the interface has to incorporate elements of knowledge-based or expert systems that are capable of as- sisting any non-expert user to select, set up, run, and interpret specialized software. By providing a coherent user interface, the interactions between different models, their da- tabases, and auxiliary software for display and analysis become transparent for the user, and a more experimental, educational style of computer use can be supported. This greatly facilitates the design and analysis of alternative policies for the management of industrial risk. ^,

An important element in the overall concept is the direct coupling of large data- bases of scientific and technical information with human expertise, of formal algorithmic methods and models with heuristics and human judgement. The expertisystems a p proach not only allows direct and interactive use of the computer, it is designed as a tightly coupled man-machine system where the vastly different data handling, analysis and judgement capabilities of man and computer are integrated into one coherent frame- work. For a fuller treatment of structure and design, and the implementation of a demonstration prototype, see Fedra (1985, 1986), Fedra et al. (1987).

1.2

A

For purposes of this model we regard the chemical industry as being divided into a number of aubsectors, each for a group of closely related chemicals. These subsectors are called Production/Distribution Areas (PDAs) because they basically comprise a network of production processes and production flows for a very specific group of chemicals. In fact, PDAs often correspond roughly to the areas of production covered by individual, large chemical companies; it makes sense for each company to deal with a particular, closely related group of chemicals because they can then coordinate the flow of inter- mediates, feedstocks, etc. through a set of linked processes with the minimum of depen- dence on external suppliers. These PDAs wish to maximize their profits by developing the most efficient production structure for a given economic, social and/or political en- vironment; since this environment is constantly changing, the production structure must evolve to keep pace with it. One very important application of the PDA model could therefore be to help in determining the best production structure for an individual com- pany u~lder various operating conditions. In addition, by adjusting the boundaries of the PDA it is possible to determine how individual companies could broaden their range of activities most effectively.

To do this, a large set of technologies is considered, mapping the input resource vector onto the production goal vector (demand vector) in a given environment. The production goal may be either based on observed data or modeled according to some scenario. Using this goal and assuming that it excludes wasteful consumption, it is possi- ble to determine the production structure for commodities that best meets this demand.

Then, working backwards, and using information on the chemical precursors of each commodity, it is possible to determine the chemical production structure that underlines the production of this combination of commodities.

Therefore the general PDA model takes into account:

the processing of flows of chemicals within the PDA,

the flow of chemicals into and out of other areas or industries, representing the marketing or business activity of the PDA,

the flow of investment, revenue and other resources such as energy, manpower etc.

More specifically, the model describes a set of possible modes of production, includ- ing alternative ranges of products made a t a given installation, recycling of semiproducts and coupled production of a number of chemicals a t one plant. Since the model is aimed at formulating decision problems concerned with the generation of efficient development alternatives for a PDA, it comprises additional constraints on resource availability as well as a set of objective functions reflecting the preferences or goals of the decision mak- er. The latter option generally leads to the formulation of a multiobjective optimization problem; the relative importance of various trade-offs arising in the problem can only be assessed by the decision maker. This is why it is important to use an interactive decision support system in conjunction with this model.

Despite the generality of the tool, real applications may be quite different for the various industrial sectors and case-specific assumptions. Decision problem formulations, scenarios, and subsequent numerical experiments, are different in many cases, so that the PDA model has to be tailor-made for each application.

The implementation of the extended and refined version developed for this case study of the pesticide industry is discussed below. The current issue of the model is described in detail in section 3 of the paper.

The PDA optimization module is a core of the interactive decision support system.

Its aim is the selection of development alternatives in the pesticide industry with special emphasis on environmental impacts. The module comprises the data on the pesticide production technologies configured according to the structure of the optimization p r o b lem based on the PDA model. The model is solved by means of a linear programming package POSTAN, developed by JSRD, which is an extension of MINOS (Stanford, 1981). A s far as multiobjective optimization experiments are concerned, the OPTIMIST package which is JSRD's enhanced version of the IIASA package MM (Kreglewski &

Lewandowski, 1983) is used.

The above software has been integrated into the overall system with advanced, user-friendly graphic display, connections to various interactively accessible background databases, an environmental impact analysis (river water quality) simulation model, and a post-processor for discrete multi-criteria optimization.

For the demonstration prototype of the system, a pesticide PDA was selected. The PDA was assembled from the set of processes used by several factories. The rorrespond- ing technological network includes some synthesis and f~rmulat~ion processes that are carried out on the basis of active substances from domestic and external (i.e., outside the PDA) production. The pesticide PDA comprises the following installations of chemi- cal syntheses:

methoxychlor,

akaritox (tetradifon), chlorfenvinphos,

chlorofos (trichlorofon; dipterex), malathion,

sodium trichloreacetate, and copper oxychloride,

out of which only the last two are pesticides which can be used directly as final products.

The products of other syntheses provide interim products for the formulation process.

The PDA includes the following formulation plants:

jet mill,

Venuleth's mixer,

active substances spread installation, formulation of liquid pesticides, condux system.

1.3

The Pesticide Industry Case Study

For the specific implementation of this interactive decision support tool for indus- trial structure optimization, a case study of a pesticide industry was chosen since many of the raw materials, products, and wastes involved are hazardous substances and thus provide connections to the overall framework system with its set of risk management and impact assessment models.

The enormous increase in the use of pesticides during the past thirty years has fueled a major controversy. Proponents argue that their use is necessary to provide an adequate food supply and t o protect human health. Opponents dispute this contention and claim severe damage is being done to our environment, with adverse effects on fish, wildlife, and, most worrisome, human health.

Worldwide consumption of pesticides was 4,571,000 metric tons in 1980. The an- nual growth rate from 1980 to 1995 is forecasted to vary between 2.3%-4.5%. This means that world production will be around 6,500,000 metric tons in 1995.

During the 1970's consumption of pesticides worldwide increased significantly, par- ticularly in the developing regions where there was an approximately 190% increase from 1970 to 1975, implying an annual growth of 30% (Ahmed, 1985). This particular growth in the use of pesticides in the developing countries leveled to about $ 1000 M/a sales by the late 1970's in constant dollar terms. Figures for 1983 showed that there has been little change in actual sales since the late seventies. One major change occurred -

the USA has become the major exporter of pesticides worldwide. Until about two or three years ago it was the second largest exporter of pesticides. The most recent figures indicatc that the USA accounts for about 29.5% of the global export market, followed by West Germany (19.5%) and the UK (12.5%). The developing countries account for about, 6% of the U S export, market and up to $ 3.8 billion worth of pesticides are export- ed. The developing countries import 40% for its needs.

Frost & Sullivan, lnc., in 1984 forecast 1.5% real annual growth in Western Europe's consurnpt,ion of pesticides lwt,ween 1982 and 1989 (Tables 1 & 2). Current sales are expected to t,ot,al $ 3.8 billion in Western Europe, rising to $ 4.1 billion in 1989 (all figures are stated in constant 1982 dollars).

"Biological control methods and Integrated Pest Management (IPM) will expand their uses with the help of persuasion and trained personneln, says the report, but the in- creasing world need for food resources is seen insuring continued growth in use of chemi- cal methods, though technological, regulation and environmental considerations will probably slow the rate of products innovation. One area of possible breakthroughs cited is development of highly specific agents through biotechnological techniques: piperidine derivatives have already been suggested in this connection.

Table 1. Pesticides production and consumption statistics i n some European countries

Production (in thousand tons)

1980 1981 1982 1983 1984 1989 1995 Annual Growth Source

BUL 40

CSSR 16

G D R 51

HUN 28

POL 9

ROM 40

USSR 474

E.Eur 723

E C 76

WORLD 4571

Consumption (in M constant 1982 US $)

FRA 1020 490 1 .O%

G F R 450 490 1.1%

4

I T A 440 490 1.4% 4

NET 90 90 0.0% 4)

SPA 2 10 270 3.6% 4)

UK 430 430 0.0% 4)

4 1) Facts 6 Figures for the Chemical Industry. Section Five: Foreign Chemical Industries. Chemical and Engineering News, June 9, 1986, Vol. 61, Nr. 23, pp.83-84.

2) Predicaets, Lnc. Worldcaats, 1984, pp.Bl14B115.

3) Food and Agricultural Organization of the United Nations. Production Yearbook, 1985, Vo1.35, p.316.

4) Frost k Sullivan, Inc. News, Report Nr. E635, May 25, 1984, p.4.

The 12 top pesticides manufacturers in order of sales volume, (listed by countries where the headquarters are situated), according t o Frost k Sullivan (1984), are Bayer (West Germany), Ciba-Geigy (Switzerland), Monsanto (USA), Royal Dutch/Shell (Netherlands/UK), Hoechst, including Roussel-Uclaf (West Germany), Rh6ne-Poulenc (France), BASF (West Germany), Schering, including FBC (West Germany), IC1 (UK), DuPont (USA), Union Carbide (USA) and Eli Lilly (USA).

Though US companies are increasingly challenging West European companies on their own turf - for instance Monsato's successful introduction of the glyphosate herbi- cide "Roundupn - Europe's dominance appears assured for the forecast period.

The Frost & Sullivan study also contains a special analysis of trade patterns in pes- ticides within Europe and between Europe and the rest of the world. It is demonstrated that West Germany is by far the biggest exporter in Europe, though Switzerland has the highest ratio of exports t o imports. France is the biggest importer.

Table

t .

197476 198 1 1982 1983

DDT

BHC (benzenehexachloride) Lindane

Aldrin and rel. insecticides Toxaphene

Other Chlor hydrocarbons Parathion

Malathion

Other org. phosph. insecticides Pyrethrum

Other horticultural insecticides Arsenicals

Carbamates insecticide Dinitro compounds Mineral oils

Other insecticides Sulphur

Lime sulphur Copper compounds Dithiocarbamates Aromatic compounds Other fungicides

Seed dress org. mercurial Seed dressings others 2.4-D

MCPA 2.4.5-T Triazines

Carbamates herbicide Urea derivatives Other herbicides Bromides

Other fumigants An ticoagulants Other roden ticides Pesticides NES

Source: F A 0 Production Yearbook, 1985.

Trade in pesticides within Western Europe is much larger than t h a t between Western Europe and the USA, which in turn is greater than trade with Japan. France, West Germany and the UK are all found to be important, and approximately equal, ex- porters to the developing countries and the East bloc.

As regards pesticides classes, Frost & Sullivan find herbicides t o have t h e largest market share but t o be growing slowest. Spain will exhibit the fastest country growth rate through 1989. End uses are dominated by agriculture, with horticulture being com- parable in consumption with industrial plus household uses.

Pesticide consumption has led t o a number of problems (Ahmed, 1985):

t h e problem of misleading marketing practices by both producers and distributors;

the use of pesticides t h a t have been banned or severely restricted in t h e exporting countries;

consumer misuse and abuse;

inadequate education and training of pesticide users and applicators.

T h e pesticide case study with the model-based decision support system described in

. this paper will not deal directly with these questions. However, it provides an impor- t a n t DSS type framework which may serve the purpose of solving these problems in a cooperative way between policy-making bodiec. and industry.

2. A Guided Tour through the Model System

T h e model system described here is implemented on a high-resolution color graph- ics workstation. Its user interface is completely menu-driven, i.e., a t any point in time, all the possible options for t h e user are indicated by a'system of hierarchical menus and associated explain functions. T o give a vivid impression of how this type of interactive model works, we will structure t h e following description along the lines of this interface;

however, since a major feature of t h e system is its conversational and arbitrary sequence style, any necessarily sequential description will be found wanting.

Their are several main components embedded in the system:

a a group of interrelated databases;

a the linear programming model with its interactive and hierarchically structured output control;

a environmental impact analysis ( a river water quality model);

a post-processor for scenario comparison;

all integrated within a uniform graphics-oriented user environment. T h e system is characterized by a high degree of connectivity, i.e., most of t h e modules, and in particu- lar the databases, can be called from various places (e.g., whenever a substance listing is on the screen), and there are numerous cross-references linking the modules.

All these connections are designed t o provide a "natural" extension of the informa- tion displayed in any one screen, such t h a t the full amount of information is available t o the user v i a one or several menus, without any of the displays being overloaded.

2.1 Getting Started

After starting the model a t the command interpreter kvel (the shell in t h e UNIX environment used), the interactive program takes over and presents the start-up screen (Figure 1).

The menu on t h e s t a r t u p page provides two standard options found in any menu of the system: S T O P and EXIT, or RETURN a t any lower level, t h a t will transfer control t o the next higher (previous) level, and EXPLAIN. T h e EXPLAIN option will darken the screen, so t h a t t h e current context is only dimly visible, and display some explanato- ry text from a database of explanation text files, referenced by the id defined for the current position, status, o r context of calling this menu option.

'I)!\$\ IDdustraI S t r u c t u r e Optlrnlzatlon P e s t ~ c l d r Industry Case St* .

I'L.\II. LJ. S y ~ t t h q * \ ~ \ mtkl ~ w u I Y I ~ ~ w I : LL.

widL.r 15

-1- d.trorr S b L t r t a l s r

-

8 U

o t b u s r ^{5 ) i}

8nte mi-

v r d d r t r l2

w l i & rat+: 7

..

rwiaratcr w

m i t e r i a tor Q p t W m t i a r rt i v t r a n l a p.r i r o r t r a n l a m t e r i a l ~ t i a -t

tram-t -1

r a l ~ y camq#tla (tech.) t a t . 1 ..ta -ioa

t a t . 1 l i v i d ...tc

-

' t

ROdDCtioa TCCLIIOIQI~CII m p t k . 1 ~ p l a t s m Jet .Ill

vcauletb .I- m

fa-mlmtloa pleat

.p.ed pleat IR

con&u system

Figure 1: Start-up page with top-level menu

2.2

Databases and Background Information

T h e start-up page menu provides access t o the three major databases of the sys- tem. These databases provide background information, in part combined with the latest results from the optimization experiments. The three databases cover:

Production Technologies;

Waste Streams and a related Industrial Establishments database;

Hazardous Substances.

2.2.1 Production Technologies

The technology database contains d a t a describing technologies that are assunled to be site independent, i.e., inputs and feedstocks, waste products and by products, trace contaminants, a qualitative hazard rating, and finally a process flow sheet or unit equip- ment layout. A cross-connection t o the hazardous substances database (see 2.2.4) uses these listings of substances for reference.

An individual process technology'can be selected from a page with process names, by picking the appropriate name with the mouse picking device. The individual process technologies are grouped by the installations they can run on (see 3 . 1 ) .

In general, database structures are designed t o accommodate a list or frame- oriented extension, with a DBMS implemented in LISP. Database structures are open- ended, t o allow the inclusion of additional information (e.g., detailed eq~~ipmerlt descrip- tions as required by the fault-tree analysis package SAFETI, Technica 1984) a t a later stage without requiring a complete redesign of the databases and mar~agc~ncnt software.

Figure f Production technology database: standard page display.

Also, there are several levels of individual process simulation and optimization available or under development in the framework system (Fedra, 1986; Grauer and Fedra, 1986;

Winkelbauer, 1987) that could be connected t o the display software for this database.

2.2.2 P r o c e s s W a s t , e S t r e a m s

The process wade stream database contains information on the 48 waste streams of the pe~t~icide model. A specific waste stream is selected from a one-page listing of all the waste streams in the system.

This information includes:

Ihe name of the waste stream, inclrlding codes and acronyms where applicable;

the number of establishments producing it, and a map displaying the individual in- st,allat,ion locat,ion with a symbol scaled t o represent production volume (since the current version of the model is spatially aggregated, allocation to individual sites is proportional, based on a fixed ratio);

total production volume (reflecting the current status of the model, or the default status quo when called before an optimization experiment);

physico-chemical waste stream characteristics such as specific gravity, water and ash content,, heatring value, BOD, biodegradation rate, etc.;

substances of conc,ern, including their mass fraction and basic physicechernical data.

--

L

-

---

-

--- ^{- ---- -}

q .

; ~ . t r ~ u ~ . t ~ ~ ~ ~ u ~ ~ t r h :!; l ~ ~ ~ / . n ~ ~ ~ ~ ~ * h ~ ~ ~ ]

, . ^{, .} .--.

'

-

' I "

- -.

7 - 1 - r

3."-

I>::..

Figure 3: Waste stream database: standard page display

T h e waste stream database can be called from various levels; in addition t o the toplevel front page, the display of results on a technology level offers another entry point. Also, the substances database, in its production information window (see 2.2.4) lists waste streams that can be used as an entry point t o the waste stream database.

2.2.3 Industrial E s t a b l i s h m e n t s Database

The map displaying the size (in terms of contributing t o the waste stream in ques- tion) and location of certain industrial installations serves as an entry t o the industrial establishments database. An establishment or plant is identified by picking it with the mouse pointing device from the map.

lnformation on a specific plant or establishment includes the name and location of the enterprise, information about its size (total production volume or number of employ- ees), a list of major product groups, production technologies, and finally a listing of sub- stances involved in the production t h a t are subject t o the E C Directive on the major ha- zards of certain industrial activities (82/501/EEC No.230).

Again, the locations database offers cross-connections t o the substances database as well as the production technologies database.

2.2.4 H a z a r d o u s S u b s t a n c e s Database

I The hazardous substances database contains information on the 125 substances of the pesticide model. They may be feedstocks, interim products, products, or wastes.

Access a t the start-up page level is either through a multi-page listing, from which a

~ y d s t w ~ . and

...-_

-

Figure

4:

substance name can be picked with the mouse pointing device, .or through a parsing function t h a t allows direct entry of substance names, or significant parts of names. T h e latter can result either in a single or in multiple references. In the first case, the corresponding database page will be displayed directly, in the latter, a list of all sub- stances referred (ranked by probability, see Fedra e t al., 1987) is displayed for further selection.

Substance listings are coded with a set. of symt)ols to denote e.g., their inclusion in specific EC regulations, high toxicity, specific water pollutants, fire and explosion ha- zard, etc. For a more detailed treatment of the hazardous substances database imple- mentation see Fedra e t al. (1987).

T h e substances database can also be accessed from various other parts of the sys- tem, whenever a listing of substance names is on the screen: selecting the appropriate menu option and selecting a substance name will display the corresponding database page-

Information on individual substances includes name, synonyms, various ID numbers (CAS, U N ) , a summary description of state, appearance, odor, solubility and persistence in qualitative terms, health-related toxicity information, including symptoms and types of exposure; chemical formula, including a color-coded representation of chem- ical structure, a table of physicechemical dat,a, reference t o legislation and regulation covering the substance, and finally production informatlion are provided. T h e produc-

.d tion information includes, in addition to average prodl~ction figures, a list of production processes involving t h e substance (eit.tler as a feedstock, interim or by-product, or as a waste product), as well as the associated waste streams. These waste streams can be

.apaccsd cuclaogm Irritative 4-L.c.~llgr.lcarl~~Lwt f.llrc -lei- ,.I& Urrt g + l e

r forrula r

rr

1 e t . I L L O C t Q I -1- t.IL bettor

e

milt f- m f f

~lrlatlonr

Dlrrtl*. C lWUer C s 6 7 n m t

Figure 5: Hazardow substances database

used as entry points t o the waste stream database described above. The production processes can be used also ^as references t o the process technology database.

2.3 Numerical Experiments: Interactive Optimization

T h e final menu option of the start-up screen will initialize the optimization model, read in the d a t a , perform an optimization run for each of the seven objectives in turn t o find the respective utopia and nadir points, and finally generate results for an uncon- strained simultaneous optimization of all seven criteria, with the utopia point serving as an implicit reference point. The results of this optimization run are displayed in numeri- cal as well as in graphical form. T h e numerical representation is simply a list of numbers of the current and previous values for the seven criteria (which in the case of the first run are the same). In addition, these values are displayed as the position of a red arrow within the interval from original upper t o lower bound for the respective cri- teria. In addition t o the current solution, the position of the relevant bounds (upper or lower, depending on whether the criteria values are minimized or maximized) is indicat- ed by a red bar. The utopia points are shown as yellow dots, and the reference points (in the first run coinciding with the utopia point) are marked by little blue arrowheads.

T h e network of installations introduced on the start-up page is displayed again.

This time, t,he boxes representing individual production technologies are filled according t o the process capacities utilized in the different installations (Figure 1).

:I(!\%\ lndustral Structurr Optimization Prsticidr l ndustry Care Study

.,I 1 : SI \: mtatm.

rc2-

9227

-1-8-t LCU maximize %la

7

23417

I ]pat.

-

t LCU minimize - = W a CiAL minimize water ..C T U S minimize.

11qld r m t e T L W minimize.

12.19

R

Figure 6: T o p level of the interactive optimization: display of results for criteria and main control menu

The control menu, in addition to the standard EXPLAIN and STOP options, con- tains three types of options:

problem formulation or scenario definition;

oulpnt display and analysis;

(re)run the problem solver

T o define a scenario for analysis, the user can manipulate the problem definition on thc. level of the criteria, by choosing to consider or ignore any subset of the criteria; he can place constraints, that is upper or lower bounds on the values of the criteria, e.g., specify a maximum amount of liquid waste; and finally, he can define a desired outcome in terms of a target or reference point, setting criteria reference values a t the levels that he would prefer them to be.

The sccorid level of problem formulation is the level of production technologies, and substances. Here the user can specify upper or lower bounds on the capacities for a pro- duction process (forcing the model t o turn the respective process on or off), he can define target amounts for production (and the corresponding prices for either selling or buying the substance), and he can finally limit the amount of any specific waste the model may produce by directly putting a constraint (upper limit) on it, or by indirectly specifying a high price for treatment (which could also represent a form of waste tax).

2.3.1 Defining Scenarios for optimization

For the problem definition a t the level of the criteria, the user can choose between the following menu options:

select optimitation criteria conetrain criteria rangee define a reference point.

!(It\%\ ladustrel Structure Optimization: Pesticide l n d u s t j Cesc Study

I

-

- "'Z -

I

1249

-<-

prths\ A M &ITTON to met v m l r .

F i g u r e 7: Setting a bound

These three menu options all directly or indirectly affect the problem definition at,

the criteria level. The model currently is implemented for seven criteria:

net income;

sales and export or gross income;

import of raw materials;

waste treatment costs;

energy use;

water consumption;

liquid waste volume.

Select Optimization Crilcria

O u t of this list, the user can specify any subset by just pointing at the criterion and toggling the s t a t u s indicator between minimize

/

/

In practice, a user could s t a r t with a small subset of the criteria t h a t he considers t o be most important. After having found a feasible solution t h a t is satisfactory in terms of this criteria subset, additional criteria (which so far have only provided auxili- ary information) can be included, one by one, t o represent additional concerns requiring further trade-offs.

Constrain Criteria Ranges

For active criteria (those t h a t have their status indicator set t o either minimize or maximize), the user can set upper or lower bounds, respectively, by dragging the con- straint indicator ( a red bar across the displayed range) with the mouse. While the con- straint is moved, the respective numerical values are indicated t o provide additional in- formation. The value is set by just clicking the appropriate mouse button, indicated by the prompt string t h a t provides additional information after a certain menu option was selected, wherever necessary.

Define a Reference Point

Finally, the user can define a reference point (other than the default utopia point) by selecting, dragging, and positioning any of the active criteria reference point symbols with the mouse. For a description of the underlying philosophy of multi-criteria optimi- zation, see Wierzbicki (1983). A formal description of the Reference Point Approach is given in section 3.3.

Another four menu options lead t o a hierarchy of display and selection levels.

These options are:

d e f i n e p r o d u c t i o n t a r g e t s ; c o n s t r a i n p r o c e s s c a p a c i t i e s ; c o n s t r a i n w a s t e p r o d u c t i o n ; edit the c o s t coefficients.

Define Production Targets

T o define production targets, the user calls up tables listing 38 main products, to- gether with their current production level, current target production level (which is the same as the current production for the first run), a percentsage range around the target, and the import and export prices lor that substance. Thc latter four values the user can change by identifying the selected number, and then modify it either with the mouse (pressing the left button will decrease the number with increasing speed until it reaches i t s allowable minimum value; pressing the right mouse button will increase the number until i t reaches its upper limit; and pressing the middle button will set the number t o t h e current value) or by direct keyboard entry.

This process can be repeated for as many numbers or as often for any number as the user chooses.

Since the model will balance any internal production deficit (due t o the various constraints on processes or wastes) by imports, setting import versus export prices can considerably affect the model.

!Ib\$\ l n d u s l r a l S l r u c l u r r O p l i r n i z s l i o n P r s l ~ r i d r I n d u s t r y Case S l u d y

I . a s ) I I*) . b l t n f i I 1% . ~ . I B ~ , I I , , ~ I I.II + t 8 - ~

,

k e y b a r d entry: 36#)

(mlnl. r l l h I F T U )

Figure 8: Defining production targets

In the worst case, it turns into a trading model, and, given stringent constraints on waste production and waste treatment prices, will resort t o importing increasing amounts.

Constrain Process Capacities

The next menu option allows one to constrain process capacities. For the 38 ma- jor production technologies of the model, a table with the current production level (in tons per year) and the upper and lower bounds around that level is displayed (Figure 9).

The constraints on the production capacities serve a dual purpose: on the one hand, they can restrict or even completely ban certain production technologies. A possible ex- ample might be high-risk technologies involving super-toxins such as various forms of furan or dioxin.

Alternatively, the lower bound can be used to force a certain minimum level of pro- duction for strategically important substances, no matter what the costs involved, or represent the situation of a subsidized industry where production capacity is held artificially high for socio-poli tical, e.g., employment reasons.

Constrain Waste Streams

An important option is the constraining of wastes. Here the model offers two different forms of control: direct constraints, and indirect economic control via waste treatment prices (which could also represent a waste tax).

Production T e c h n o l ~ i v s and I'IXXI

\ \ , l l i l . . . I . ^!.I.,,,!

d l a a l 0

'

m b l a 0

r L r I t m ( t e t r d l f o a ) 0

d l a f a l m f o r @

u l a t i o n (fast ion)

f d l a ( t r l c b l a f m ) 0 c q q ~ ~ a x y c b l a i h 0

1 \ 3 1 , 1 1 1 l I . I I I.'II 1 ' 1 11111

.&for 34 raoraw I prapotox r t o x 34

=-cp

r t o f o r 3

u l a s g 40 f o r c h l a 25 m o l o f o s 50 Iasochroo d q l ~ i k

1000 2000 2000 2.00 2060 2500 2000 0 2500

e

2000 3000

moo moo moo moo moo

1500 1000

, , . , /I I , . , . I , . , ! 8 , , (I

Current capacity: 3112.00 tonsla . .

W r bound: :,-160. 011

Lower bound:

keyboard emtry: 300

It-lnl. -4th ItiU)

p r r - s s I I I 1 14 1111'. I ~1 111 hI ".I , L I 1 1 1 1:I I la^',

I..

Figure 9: Setting constraints on process capacities

For the 40 major substances in liquid waste streams, a table which includes sub- stance name, current level of generation, maximum allowable level (upper bound), and a treatment price (and/or waste tax) per ton of product is displayed. Again the latter two values can be set for each substance by mouse buttons or by direct keyboard entry.

Edit Cost Coeficients

Finally, a few global cost coefficients (unit cost of labor, energy prices for electrical and technological (steam) energy, price of water, etc. can be set, in particular t o analyze the system's reaction t o changes in labor costs and energy. However, since a change in energy prices would in all likelihood also trigger a change of the p r o d ~ ~ c t prices (defined above), it would be extremely difficult t o formulate a consistent scenario includ- ing massive changes in energy prices. Currently, no mechariisnl other than simple upper and lower bounds on the values the user can change are implemented t o ensure scenario consistency. This, however, would be a challenging extension of the current model.

2.3.2 Running the Problem Solver

After a scenario for analysis has been defined, the menu option run t h e problem solver will rerun t h e optimization package with the newly specified objectives, con- straints, and coefficients.

Figure 10: Setting constraints on waste products

The new results will be indicated by a shifting of the red results arrows, and the updating of the corresponding numbers (compare Figure 6).

If however, the user has made excessive changes t o the scenario definition, or defined too narrow or even contradictory constraints, no feasible solution can be found.

The problem solver will indicate this with a short diagnostic message, after which the main control menu for scenario definition will be offered again. If the user has made only one or a few changes since Lhe last feasible run, the necessary modifications of the scenario definition will be obvious and easy. As an emergency escape, however, the user can restart the entire process, re-loading the default data, and thus overwrite the current (infeasible) status of the system. All completed interim solutions, however, will still be available for analysis with the post.-processor (see section 2.6).

2.4 Model Output and Evaluation

The top level for the interactive optimization displays the current results for the seven criteria, and the values from the last run in numerical terms as well as graphical representation (Figure 6 ) . It also indicates the position of bounds, utopia point, and reference point in graphical form.

The menu option d i s p l a y r e s u l t s will call up a more detailed representation of a run's results, together with a new menu for further control of the hierarchically organ- ized model output.

The screen summarizes the results a t the industry level. A table with a breakdown of the industries' economics is provided together with a pie chart, in parallel, summariz- ing the major cost components. In addition, basic resource consumption and summary

ilk\%\ Industral Structure Optimization. Peslicidr Industry Case Study

I

Fa-,-ll,ml*~\:

d m rd toul Y C U I*13@.442 rd 1- t o t a l YCL' -14901.140

rz=-

C.rr of ..tar tt0t.l) Y C U -0.037 W i t . 1 m rd & p a l a t l a Y C U -578.000 -t t o t a l t a t YC1' -49.61 b bt m i cp r f a v w c Y C U I / ) l . ? l i .

m P. II , . I I , ~ ) tmlb~-t.l % t b f I ' I I I ~ I I I ~ I ~ ~ U I I I ~ . I ,.:

d h a a l K tans 10.344

b I at- K tons 1.500

f d l a 15 K t m s 2.000

1

- d K t ~ n s 2.000

r f i.pol4mrbmtor K tons 4.000

dm I K tons 4.000

.Ed L s i . 0 ^{1; tons} 1.500

Iprobct sales .od a p o r t s l l i q o r t s o f r n u t e r i m l

I

Gross i n c m - L d w r cost

h . ' . . P I , . 1 l ~ l l ' , ! l l l l , ~ l 1 1 1 1 1 :

energy use ( e l e c t r i c a l ) Wh 28579.8#2 energy use (technological) Wh 1 U . m rmter c o n s l q t i o n K d 3707.235

Wmter c o n s l q t ion: 1

Wmste rmter:

S o l i d r a s t e r Wseous r a s t e s :

Figure 11: Display of general results and menu for the detailed results output waste production is indicated in numeric as well as graphical form.

We also display a listing of those products, that were imported (at least in part) in order to fulfill the required target production levels.

The output control menu contains the following options:

r e s u l t s by technology;

p r o d u c t sales a n d e x p o r t s ; i m p o r t s a n d r a w m a t e r i a l s ; process w a s t e : l i q u i d s ; process w a s t e : solid;

process w a s t e : g a s / d u s t . 2.4.1. R e s u l t s by Technology

This option calls up the table of production technologies, grouped by installations (see Figure 9). The user can select any of the technologies by picking it with the mouse, and either retrieve the corresponding technologies database page for background infor- mation, or call up a screen with the current optimization results summarized for this technology (Figure 12).

The page indicates the technology chosen and the current production level together with an indication of the relative capacity utilization. i t also provides the same econom- ic breakdown as the overall results page, including more detailed production costs.